rdelaage
/
postgresql
mirror of https://git.postgresql.org/git/postgresql.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Last-minute updates for release notes. 

Security: CVE-2017-7546, CVE-2017-7547, CVE-2017-7548
This commit is contained in:
Tom Lane 2017-08-07 11:46:20 -04:00
parent fca17a933b
commit a8b37ebe40
5 changed files with 707 additions and 355 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

199
doc/src/sgml/release-9.2.sgml
View File

@ -29,7 +29,12 @@
</para>
<para>
However, if you are upgrading from a version earlier than 9.2.20,
However, if you use foreign data servers that make use of user
passwords for authentication, see the first changelog entry below.
</para>
<para>
Also, if you are upgrading from a version earlier than 9.2.20,
see <xref linkend="release-9-2-20">.
</para>
@ -40,6 +45,126 @@
<itemizedlist>
<listitem>
<para>
Further restrict visibility
of <structname>pg_user_mappings</>.<structfield>umoptions</>, to
protect passwords stored as user mapping options
(Noah Misch)
</para>
<para>
The fix for CVE-2017-7486 was incorrect: it allowed a user
to see the options in her own user mapping, even if she did not
have <literal>USAGE</> permission on the associated foreign server.
Such options might include a password that had been provided by the
server owner rather than the user herself.
Since <structname>information_schema.user_mapping_options</> does not
show the options in such cases, <structname>pg_user_mappings</>
should not either.
(CVE-2017-7547)
</para>
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user
AND (pg_has_role(S.srvowner, 'USAGE')
OR has_server_privilege(S.oid, 'USAGE')))
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>
<para>
Disallow empty passwords in all password-based authentication methods
(Heikki Linnakangas)
</para>
<para>
<application>libpq</> ignores empty password specifications, and does
not transmit them to the server. So, if a user's password has been
set to the empty string, it's impossible to log in with that password
via <application>psql</> or other <application>libpq</>-based
clients. An administrator might therefore believe that setting the
password to empty is equivalent to disabling password login.
However, with a modified or non-<application>libpq</>-based client,
logging in could be possible, depending on which authentication
method is configured. In particular the most common
method, <literal>md5</>, accepted empty passwords.
Change the server to reject empty passwords in all cases.
(CVE-2017-7546)
</para>
</listitem>
<listitem>
<para>
On Windows, retry process creation if we fail to reserve the address
@ -410,77 +535,9 @@
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
follow the corrected procedure shown in the changelog entry for
CVE-2017-7547, in <xref linkend="release-9-2-22">.
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user)
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>

199
doc/src/sgml/release-9.3.sgml
View File

@ -23,7 +23,12 @@
</para>
<para>
However, if you are upgrading from a version earlier than 9.3.16,
However, if you use foreign data servers that make use of user
passwords for authentication, see the first changelog entry below.
</para>
<para>
Also, if you are upgrading from a version earlier than 9.3.16,
see <xref linkend="release-9-3-16">.
</para>
@ -34,6 +39,126 @@
<itemizedlist>
<listitem>
<para>
Further restrict visibility
of <structname>pg_user_mappings</>.<structfield>umoptions</>, to
protect passwords stored as user mapping options
(Noah Misch)
</para>
<para>
The fix for CVE-2017-7486 was incorrect: it allowed a user
to see the options in her own user mapping, even if she did not
have <literal>USAGE</> permission on the associated foreign server.
Such options might include a password that had been provided by the
server owner rather than the user herself.
Since <structname>information_schema.user_mapping_options</> does not
show the options in such cases, <structname>pg_user_mappings</>
should not either.
(CVE-2017-7547)
</para>
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user
AND (pg_has_role(S.srvowner, 'USAGE')
OR has_server_privilege(S.oid, 'USAGE')))
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>
<para>
Disallow empty passwords in all password-based authentication methods
(Heikki Linnakangas)
</para>
<para>
<application>libpq</> ignores empty password specifications, and does
not transmit them to the server. So, if a user's password has been
set to the empty string, it's impossible to log in with that password
via <application>psql</> or other <application>libpq</>-based
clients. An administrator might therefore believe that setting the
password to empty is equivalent to disabling password login.
However, with a modified or non-<application>libpq</>-based client,
logging in could be possible, depending on which authentication
method is configured. In particular the most common
method, <literal>md5</>, accepted empty passwords.
Change the server to reject empty passwords in all cases.
(CVE-2017-7546)
</para>
</listitem>
<listitem>
<para>
Fix concurrent locking of tuple update chains (&Aacute;lvaro Herrera)
@ -497,77 +622,9 @@
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
follow the corrected procedure shown in the changelog entry for
CVE-2017-7547, in <xref linkend="release-9-3-18">.
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user)
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>

213
doc/src/sgml/release-9.4.sgml
View File

@ -23,7 +23,12 @@
</para>
<para>
However, if you are upgrading from a version earlier than 9.4.12,
However, if you use foreign data servers that make use of user
passwords for authentication, see the first changelog entry below.
</para>
<para>
Also, if you are upgrading from a version earlier than 9.4.12,
see <xref linkend="release-9-4-12">.
</para>
</sect2>
@ -33,6 +38,140 @@
<itemizedlist>
<listitem>
<para>
Further restrict visibility
of <structname>pg_user_mappings</>.<structfield>umoptions</>, to
protect passwords stored as user mapping options
(Noah Misch)
</para>
<para>
The fix for CVE-2017-7486 was incorrect: it allowed a user
to see the options in her own user mapping, even if she did not
have <literal>USAGE</> permission on the associated foreign server.
Such options might include a password that had been provided by the
server owner rather than the user herself.
Since <structname>information_schema.user_mapping_options</> does not
show the options in such cases, <structname>pg_user_mappings</>
should not either.
(CVE-2017-7547)
</para>
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user
AND (pg_has_role(S.srvowner, 'USAGE')
OR has_server_privilege(S.oid, 'USAGE')))
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>
<para>
Disallow empty passwords in all password-based authentication methods
(Heikki Linnakangas)
</para>
<para>
<application>libpq</> ignores empty password specifications, and does
not transmit them to the server. So, if a user's password has been
set to the empty string, it's impossible to log in with that password
via <application>psql</> or other <application>libpq</>-based
clients. An administrator might therefore believe that setting the
password to empty is equivalent to disabling password login.
However, with a modified or non-<application>libpq</>-based client,
logging in could be possible, depending on which authentication
method is configured. In particular the most common
method, <literal>md5</>, accepted empty passwords.
Change the server to reject empty passwords in all cases.
(CVE-2017-7546)
</para>
</listitem>
<listitem>
<para>
Make <function>lo_put()</> check for <literal>UPDATE</> privilege on
the target large object (Tom Lane, Michael Paquier)
</para>
<para>
<function>lo_put()</> should surely require the same permissions
as <function>lowrite()</>, but the check was missing, allowing any
user to change the data in a large object.
(CVE-2017-7548)
</para>
</listitem>
<listitem>
<para>
Fix concurrent locking of tuple update chains (&Aacute;lvaro Herrera)
@ -601,77 +740,9 @@ Branch: REL9_4_STABLE [23a2b818f] 2017-08-05 14:56:40 -0700
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
follow the corrected procedure shown in the changelog entry for
CVE-2017-7547, in <xref linkend="release-9-4-13">.
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user)
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>

213
doc/src/sgml/release-9.5.sgml
View File

@ -23,7 +23,12 @@
</para>
<para>
However, if you are upgrading from a version earlier than 9.5.7,
However, if you use foreign data servers that make use of user
passwords for authentication, see the first changelog entry below.
</para>
<para>
Also, if you are upgrading from a version earlier than 9.5.7,
see <xref linkend="release-9-5-7">.
</para>
</sect2>
@ -33,6 +38,140 @@
<itemizedlist>
<listitem>
<para>
Further restrict visibility
of <structname>pg_user_mappings</>.<structfield>umoptions</>, to
protect passwords stored as user mapping options
(Noah Misch)
</para>
<para>
The fix for CVE-2017-7486 was incorrect: it allowed a user
to see the options in her own user mapping, even if she did not
have <literal>USAGE</> permission on the associated foreign server.
Such options might include a password that had been provided by the
server owner rather than the user herself.
Since <structname>information_schema.user_mapping_options</> does not
show the options in such cases, <structname>pg_user_mappings</>
should not either.
(CVE-2017-7547)
</para>
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user
AND (pg_has_role(S.srvowner, 'USAGE')
OR has_server_privilege(S.oid, 'USAGE')))
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>
<para>
Disallow empty passwords in all password-based authentication methods
(Heikki Linnakangas)
</para>
<para>
<application>libpq</> ignores empty password specifications, and does
not transmit them to the server. So, if a user's password has been
set to the empty string, it's impossible to log in with that password
via <application>psql</> or other <application>libpq</>-based
clients. An administrator might therefore believe that setting the
password to empty is equivalent to disabling password login.
However, with a modified or non-<application>libpq</>-based client,
logging in could be possible, depending on which authentication
method is configured. In particular the most common
method, <literal>md5</>, accepted empty passwords.
Change the server to reject empty passwords in all cases.
(CVE-2017-7546)
</para>
</listitem>
<listitem>
<para>
Make <function>lo_put()</> check for <literal>UPDATE</> privilege on
the target large object (Tom Lane, Michael Paquier)
</para>
<para>
<function>lo_put()</> should surely require the same permissions
as <function>lowrite()</>, but the check was missing, allowing any
user to change the data in a large object.
(CVE-2017-7548)
</para>
</listitem>
<listitem>
<para>
Correct the documentation about the process for upgrading standby
@ -635,77 +774,9 @@ Branch: REL9_2_STABLE [1188b9b2c] 2017-08-02 15:07:21 -0400
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
follow the corrected procedure shown in the changelog entry for
CVE-2017-7547, in <xref linkend="release-9-5-8">.
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user)
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>

238
doc/src/sgml/release-9.6.sgml
View File

@ -23,7 +23,12 @@
</para>
<para>
However, if you are upgrading from a version earlier than 9.6.3,
However, if you use foreign data servers that make use of user
passwords for authentication, see the first changelog entry below.
</para>
<para>
Also, if you are upgrading from a version earlier than 9.6.3,
see <xref linkend="release-9-6-3">.
</para>
</sect2>
@ -35,6 +40,165 @@
<listitem>
<!--
Author: Noah Misch <noah@leadboat.com>
Branch: master [e568e1eee] 2017-08-07 07:09:28 -0700
Branch: REL9_6_STABLE [156099630] 2017-08-07 07:09:31 -0700
Branch: REL9_5_STABLE [36f9f6095] 2017-08-07 07:09:31 -0700
Branch: REL9_4_STABLE [b6e39ca92] 2017-08-07 07:09:31 -0700
Branch: REL9_3_STABLE [5e8e00914] 2017-08-07 07:09:31 -0700
Branch: REL9_2_STABLE [e255e97a2] 2017-08-07 07:09:32 -0700
-->
<para>
Further restrict visibility
of <structname>pg_user_mappings</>.<structfield>umoptions</>, to
protect passwords stored as user mapping options
(Noah Misch)
</para>
<para>
The fix for CVE-2017-7486 was incorrect: it allowed a user
to see the options in her own user mapping, even if she did not
have <literal>USAGE</> permission on the associated foreign server.
Such options might include a password that had been provided by the
server owner rather than the user herself.
Since <structname>information_schema.user_mapping_options</> does not
show the options in such cases, <structname>pg_user_mappings</>
should not either.
(CVE-2017-7547)
</para>
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user
AND (pg_has_role(S.srvowner, 'USAGE')
OR has_server_privilege(S.oid, 'USAGE')))
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>
<!--
Author: Heikki Linnakangas <heikki.linnakangas@iki.fi>
Branch: master [bf6b9e944] 2017-08-07 17:03:42 +0300
Branch: REL9_6_STABLE [f6fc72cb6] 2017-08-07 17:03:49 +0300
Branch: REL9_5_STABLE [127835ddf] 2017-08-07 17:04:00 +0300
Branch: REL9_4_STABLE [d5d46d99b] 2017-08-07 17:04:07 +0300
Branch: REL9_3_STABLE [b2f833ea7] 2017-08-07 17:04:12 +0300
Branch: REL9_2_STABLE [06651648a] 2017-08-07 17:04:17 +0300
-->
<para>
Disallow empty passwords in all password-based authentication methods
(Heikki Linnakangas)
</para>
<para>
<application>libpq</> ignores empty password specifications, and does
not transmit them to the server. So, if a user's password has been
set to the empty string, it's impossible to log in with that password
via <application>psql</> or other <application>libpq</>-based
clients. An administrator might therefore believe that setting the
password to empty is equivalent to disabling password login.
However, with a modified or non-<application>libpq</>-based client,
logging in could be possible, depending on which authentication
method is configured. In particular the most common
method, <literal>md5</>, accepted empty passwords.
Change the server to reject empty passwords in all cases.
(CVE-2017-7546)
</para>
</listitem>
<listitem>
<!--
Author: Tom Lane <tgl@sss.pgh.pa.us>
Branch: master [8d9881911] 2017-08-07 10:19:19 -0400
Branch: REL9_6_STABLE [52a414387] 2017-08-07 10:19:20 -0400
Branch: REL9_5_STABLE [873741c68] 2017-08-07 10:19:21 -0400
Branch: REL9_4_STABLE [f1cda6d6c] 2017-08-07 10:19:22 -0400
-->
<para>
Make <function>lo_put()</> check for <literal>UPDATE</> privilege on
the target large object (Tom Lane, Michael Paquier)
</para>
<para>
<function>lo_put()</> should surely require the same permissions
as <function>lowrite()</>, but the check was missing, allowing any
user to change the data in a large object.
(CVE-2017-7548)
</para>
</listitem>
<listitem>
<!--
Author: Bruce Momjian <bruce@momjian.us>
Branch: master [0f33a719f] 2017-06-15 12:30:02 -0400
Branch: REL9_6_STABLE [a0873fbab] 2017-06-15 12:30:02 -0400
@ -1193,77 +1357,9 @@ Branch: REL9_2_STABLE [99cbb0bd9] 2017-05-08 07:24:28 -0700
<para>
By itself, this patch will only fix the behavior in newly initdb'd
databases. If you wish to apply this change in an existing database,
you will need to do the following:
follow the corrected procedure shown in the changelog entry for
CVE-2017-7547, in <xref linkend="release-9-6-4">.
</para>
<procedure>
<step>
<para>
Restart the postmaster after adding <literal>allow_system_table_mods
= true</> to <filename>postgresql.conf</>. (In versions
supporting <command>ALTER SYSTEM</>, you can use that to make the
configuration change, but you'll still need a restart.)
</para>
</step>
<step>
<para>
In <emphasis>each</> database of the cluster,
run the following commands as superuser:
<programlisting>
SET search_path = pg_catalog;
CREATE OR REPLACE VIEW pg_user_mappings AS
SELECT
U.oid AS umid,
S.oid AS srvid,
S.srvname AS srvname,
U.umuser AS umuser,
CASE WHEN U.umuser = 0 THEN
'public'
ELSE
A.rolname
END AS usename,
CASE WHEN (U.umuser &lt;&gt; 0 AND A.rolname = current_user)
OR (U.umuser = 0 AND pg_has_role(S.srvowner, 'USAGE'))
OR (SELECT rolsuper FROM pg_authid WHERE rolname = current_user)
THEN U.umoptions
ELSE NULL END AS umoptions
FROM pg_user_mapping U
LEFT JOIN pg_authid A ON (A.oid = U.umuser) JOIN
pg_foreign_server S ON (U.umserver = S.oid);
</programlisting>
</para>
</step>
<step>
<para>
Do not forget to include the <literal>template0</>
and <literal>template1</> databases, or the vulnerability will still
exist in databases you create later. To fix <literal>template0</>,
you'll need to temporarily make it accept connections.
In <productname>PostgreSQL</> 9.5 and later, you can use
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS true;
</programlisting>
and then after fixing <literal>template0</>, undo that with
<programlisting>
ALTER DATABASE template0 WITH ALLOW_CONNECTIONS false;
</programlisting>
In prior versions, instead use
<programlisting>
UPDATE pg_database SET datallowconn = true WHERE datname = 'template0';
UPDATE pg_database SET datallowconn = false WHERE datname = 'template0';
</programlisting>
</para>
</step>
<step>
<para>
Finally, remove the <literal>allow_system_table_mods</> configuration
setting, and again restart the postmaster.
</para>
</step>
</procedure>
</listitem>
<listitem>